Aptina DR-Pix Technology White Paper

Leveraging Dynamic Response Pixel Technology
to Optimize Inter-scene Dynamic Range
An AptinaTM Technology White Paper
Introduction
Real-world cameras often encounter both very bright
scenes and very dark scenes. To capture these scenes,
image sensors are typically optimized for one extreme
at the price of degraded performance for the other. The
challenge is how to design an image sensor to work
optimally in all scene conditions.
For the image sensor within these cameras, this translates
into requirements for larger full-well (FW) capacities
as well as increased sensitivity. This is true for both the
larger pixels heavily used in PCs, gaming, automotive,
surveillance, and digital single-lens reflex (DSLR) cameras
and the ever-shrinking pixel sizes used in mobile phone
applications and digital still cameras (DSCs). However,
higher sensitivity could limit FW capacity, the maximum
achievable signal-to-noise ratio (SNR), and the total
dynamic range (DR) of the sensor.
Pixels greater than 2 microns often have their FW capacity
defined by the photodiode’s charge holding capacity,
rather than the pixel’s voltage swing, due to its larger
photosensitive area. To increase the charge handling
capacity of this pixel, it is common to connect a physical
capacitor to the floating diffusion (FD) node. However, this
typically results in lower conversion gain (CG). This, in turn,
means reduced sensitivity and increased input-referred
read noise, thereby compromising low-light sensitivity
and reducing the sensor’s DR even though the sensor is
capable of measuring larger signals.
Many approaches to increasing DR focus on achieving
high intra-scene dynamic range. However, these HDR
techniques do not improve low-light sensitivity or reduce
noise to improve low-light image captures. As a result,
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved. 9/10 EN.L
a different approach has emerged to improve interscene DR and sensor performance through the addition
of a high-sensitivity mode to the sensor’s operation.
Two modes are combined in one pixel design – low
conversion gain (LCG) for large charge handling capacity
in bright scenes and a high conversion gain (HCG) mode
with increased sensitivity and low read noise for lowlight scenes, providing tremendous benefit for DSLR,
surveillance, and notebook/PC cameras, as well as
automotive imaging systems, where image sensors are
expected to capture images/video in extreme low-light
conditions without sacrificing the performance in highlight conditions to do so.
Relationship between Conversion Gain and
Full-Well Capacity
To understand the benefit of adding an HCG mode, it
is important to discuss the trade-off between CG, the
measure of the sensitivity of the charge detection at the
FD node in a CMOS image sensor pixel, and FW in a pixel
design.
The CG is actually an inverse way of expressing the
capacitance of the FD node. Capacitance can be calculated
as the ratio of the amount of charge required to change
the potential of a node by one volt (see Equation 1),
where Q is charge in Coulombs, and V is the potential in
Volts. This means that, as the capacitance of the FD node
increases, the conversion gain and hence, sensitivity of the
FD node decreases.
C= Q/V
Equation 1: Basic formula for capacitance
Leveraging DR-Pix Technology
Generally, the amount of voltage swing allowed in the pixel
is fixed by the overall sensor design. This fixed range of
voltage means that CG can have a distinct impact on the
sensor’s FW. For example, in a typical design, the power
supply is fixed at 2.8 V and the analog signal chain has a
fixed window of 1 V allowed voltage swing at the input. For
example, if 1 V of swing is allowed at the pixel’s output and
the source follower (SF) gain is 0.8, then the allowed voltage
swing at the FD is equal to 1 V / 0.8 = 1.25 V.
which is due to both parasitics (pn junction, metal coupling,
gate overlap, etc.) and possibly also the addition of a
physical capacitor (poly-insulator-poly [PIP], metal-insulatormetal [MIM], metal-oxide-semiconductor [MOS], etc.).
As shown above, the capacitance of the FD node will
determine the amount of charge that can be detected
within the fixed voltage operating range. Assuming a 1.25
FD voltage swing: if a sensor has CG = 30 μV/e, then the
largest FW capacity achievable will be 1.25∙106 μV / 30 μV/
e ~ 42,000 electrons. Alternatively, if a pixel has a much
higher CG equal to 150 μV/e, then the FW capacity would be
limited by the fixed voltage swing:
1.25∙106 μV / 150 μV/e ~ 8,000 electrons.
Importance of both Conversion Gain and FullWell
Capacitance is connected in parallel with the FD node,
therefore the addition of more capacitance means
more charge can be held, but at the expense of reduced
sensitivity.
Higher CG means higher sensitivity, as one signal electron
can be more easily detected. An example curve of signal, in
units of mV at the FD node, versus light exposure in units of
lux*s, is shown in Figure 2. Responsivity, which is a measure
of the sensitivity of an image sensor, is defined as the slope
of signal change versus change in exposure.
2000
Transfer
Gate
(TX)
1800
1600
Reset
FD
VAA_PIX
FD Signal (mV)
1400
1200
1000
800
600
PD
CFD
SF
Low CG
400
High CG
200
0
RS
VOUT
RL
Figure 1: Pixel schematic.
Figure 1, a schematic diagram of a standard 4T CMOS
image sensor pixel circuit, illustrates the role of CG in the
typical pixel design. The definitions of the labels in this pixel
schematic are as follows: PD = photodiode, TX = transfer
transistor, RST = reset transistor, SF = source follower, RS =
row select transistor, Vaa _PIX = analog pixel power supply
voltage, Vout= pixel output voltage node, FD = floating
diffusion node, and Cfd = capacitance at the FD node –
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved. 9/10 EN.L
0.0
0.2
0.4
0.6
Exposure (lux * s)
0.8
1.0
Figure 2: FD signal in mV as a function of light exposure in
lux*s for two sensors with low and high CG, respectively.
In this example, the high-CG sensor has CG=127 μV/e and
full-well = 10,500 electrons. The low-CG sensor has 30.8
μV/e and full-well = 42,000 electrons. In both cases, the
overall maximum change in voltage at the FD is nearly
equal, at approximately 1.3 V, using only the linear segment
of the signal response curve. The high-CG sensor shows a
very clear responsivity advantage; however, the trade off
is clearly seen in the much lower full-well capacity as the
sensor’s linear range ends at a much lower exposure.
Higher CG also means that the sensor will realize a
reduction in read noise when the system noise is referred
back to the FD. It is important to refer the baseline noise
of the imaging system back to the signal input, in order to
Leveraging DR-Pix Technology
compare the amount of read noise to the amount of signal
that is being measured.
same way. However, by increasing CG, the amount of fullwell achievable will be limited.
For example, if the analog signal chain of an image sensor
has 100 μV of noise at the input, then this amount of noise
can be referred back to the FD by using the SF gain. For
example, 100 μV of noise at the pixel’s output, divided by
SF gain of 0.8, becomes 125 μV of noise when referred to
the FD. Now, when divided by CG to convert to noise in
units of electrons, this shows that higher CG will result in
fewer electrons of noise at the FD. In the example above,
the low-CG case saw 125 μV / 30 μV/e = 4.2 electrons of
noise whereas 125 μV / 150 μV/e = 0.8 electrons of noise
was evident in the high-CG case. The higher CG produces
not only a larger voltage signal for a given amount of signal
electrons, it also makes the noise of the system appear
smaller when compared to the measured signal.
Expanding Dynamic Range and Low-Light
Performance
Certainly, a balance exists between a sensor’s CG and FW.
But, how do these parameters affect the max SNR and DR,
and what may be done to improve those characteristics?
How CG and FW Capacity Relate to DR and SNR
Two of the most important parameters of image sensors are
the signal-to-noise ratio (SNR) and the dynamic range (DR).
SNR is a very broad way to determine image sensor quality.
For a given imaging scene and exposure, a higher SNR value
will result better image quality. The SNRMAX (see Equation
2) is a logarithmic ratio of signal electrons at the sensor’s
saturation exposure (NMAX) to either the total or temporal
noise in units of electrons. The result is expressed in units of
dB.
Although picture resolution is increasing in today’s camera
systems, the DR of pixels continues to decrease as pixel
sizes shrink, thereby limiting the ability to produce natural
photos with both highlights and shadows preserved. As
a result, various pixel schemes have been proposed for
achieving high dynamic range (HDR), including logarithmic
pixels, lateral overflow, frame multi-exposure (ME)and
intra-frame multi-exposure (IFME), among others.
These methods of achieving HDR are aimed at achieving
high intra-scene DR, which means that dark and bright
areas within one scene can be properly exposed. This
is commonly achieved either by combining multiple
exposure times or frame captures into one image using
the multi-frame approach (example signal response shown
in Figure 3) [1,2], or by reducing the pixels’ responsivity
at higher exposures using a nonlinear signal approach
(example signal responses in Figure 4).
Q
Qsat
Highest Light
Moderate Light
SNRMAX=20∙log10 (NMAX/n)
Equation 2: Formula for signal-to-noise ratio (SNR)
Low Light
Alternatively, DR is a measure of ratio of the highest and
lowest possible signals that can be measured by the sensor
(see Equation 3).
DR=20∙log10 (NMAX/nread )
Equation 3: Formula for dynamic range (DR)
The read noise is the lowest measureable signal, and is
defined by the base noise level of the sensor’s whole signal
chain / system. The DR result is also expressed in units of dB.
A larger full-well increases the maximum achievable SNR by
increasing NMAX. It also increases the overall sensor DR in the
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved. 9/10 EN.L
tint
a
tint
t
Figure 3: Plot of signal response from high, medium, and
low light portions of a scene. ME approaches combine
two or more signal responses at different exposure times
into one image.
Leveraging DR-Pix Technology
pixel.
Q
When imaging in high light conditions, the DCG switch
is turned on, connecting the physical capacitor to the
FD node. In this way, the large capacitance of the FD
node is used to enable an LCG mode, which can handle
a large amount of signal charge.
Transfer
Gate
(TX)
Reset
Exposure
Figure 4: Nonlinear signal response vs. exposure curve for an
example logarithmic image sensor.
These techniques can typically achieve very high DR of
>100 dB within one image, which is very useful in many
applications, such as a surveillance camera properly exposing
a criminal’s face while sunlight or glare is dominating the rest
of the scene or an automotive sign recognition system, which
must read sign or road details in the face of headlights or
sunlight.
High intra-scene DR has some drawbacks as well. Nonlinear pixels may have poor color reproduction and lowlight sensitivity, as well as higher fixed pattern noise (FPN).
ME techniques not only require extra memory and post
processing, but they can have reduced SNR near exposure
transition points and motion artifacts caused by exposures
captured in a time sequence, thus requiring extra motion
compensation and very high frame rates to be useful in video
applications.
But, most importantly, what these HDR techniques do not
do is improve low-light sensitivity or reduce noise to improve
low-light image captures. Applications such as high-end DSLR
and mirrorless DSLR cameras require a large FW, responsivity
for achieving low ISO-speed conditions and very low noise for
high ISO-speed settings used in low-light conditions.
Dynamic Response Pixel Technology
As a result, a different approach, while not a competitor to
HDR technologies, has emerged. This technique, focused on
inter-scene DR, adds a separate high sensitivity mode to a
pixel that already has a large charge handling capacity.
This approach, called dynamic response pixel technology by
Aptina, or AptinaTM DR-PixTM technology, entails controlled
switching on and off of a capacitor connected to the pixel’s
FD node (Figure 5). To perform this switching, one transistor,
called the dual conversion gain (DCG) switch, is added to the
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved. 9/10 EN.L
FD
VAA_PIX
PD
SF
DCG
RS
VOUT
RL
Figure 5: Schematic diagram of Aptina’s DR-Pix
technology.
In low-light conditions, the DCG signal is turned off,
disconnecting the cap from the FD, and enabling an
HCG mode, which can be used as an extra analog gain,
inside the pixel. In this case, the FD capacitance is only
due to the parasitic capacitance of the FD’s pn junction
diffusion and metal coupling, which is much smaller
than that due to the physical cap structure. The much
lower FD capacitance results in much higher conversion
gain, higher sensitivity, and reduced read noise, at the
expense of lower maximum charge handling capacity.
This scheme requires an added transistor and capacitor
in the pixel layout, which can eat up real estate that
could otherwise be utilized to increase the photodiode
size. It also adds metal-line routing for the DCG signal
control, which reduces the size of the aperature
above the photodiode. These drawbacks create extra
challenges as pixels shrink beyond the sub-2 micron
mark. Also, this scheme is most useful when the
photodiode (PD) size is large enough to be the defining
factor for the FW capacity. When the amount of charge
held in the PD is so large that it cannot be measured
within the pixel’s allowed voltage swing, then extra
Leveraging DR-Pix Technology
capacitance must be added to the FD in order to measure
all of the PD’s charge. In this case the DCG feature can
be added to the pixel design to add the HCG mode for
improved low-light performance.
This scheme allows very high inter-scene DR by providing
high sensitivity and low read noise for dark scenes, and
large charge handling capacity for well-lit scenes -- all
inside of one pixel design. In this approach, the DR of one
scene is not increased, but the range of illumination over
which the sensor may be used is extended at the lowlight end by the addition of the HCG mode. For DSLR
applications, this means shooting high ISO without the
need for unnatural-looking flash. For surveillance and PCcamera applications that must capture video often in
dimly lit conditions, this is also a tremendous benefit.
The following images are taken from Aptina’s MT9H004
sensor, which is an APS-C sized sensor that employs Aptina
DR-Pix technology for high-end DSLR and mirrorless
camera applications. The sensor has a max SNR around
47 dB, and inter-scene DR equal to 82.9 dB due to the low
read noise of the HCG mode. In comparison, without the
HCG mode of operation, the sensor’s DR would be limited
to 69 dB of intra-scene DR in the LCG mode.
Figure 8: Aptina’s MT9H004 sensor with Aptina
DR-Pix technology with the HCG mode enabled,
resulting in higher sensitivity and lower noise. Seen in
144 lux daylight illumination and 125 ms exposure.
Figure 9: Zoomed-in portion of the HCG image taken
at low light, showing the increased sensitivity and low
noise.
When compared with the underexposed images seen in
Figures 6 and 7, the improved sensitivity and low noise of
the HCG mode under low-light conditions is evident in
Figures 8 and 9.
Figure 6: Aptina MT9H004 sensor with Aptina
DR-Pix technology, 144 lux daylight illumination,
125 ms exposure, in LCG mode.
Figure 7: Zoomed-in portion of low conversion gain image
taken in low-light resulting in underexposure.
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved. 9/10 EN.L
Figure 10: Overexposed scene using HCG in a 2,430 lux
daylight illumination scene with 125 ms of exposure
time.
An
ObjectiveDR-Pix
Look at
FSI and BSI
Leveraging
Technology
In bright scenes, large pixel capacity is required. Figures 10
and 11 display the effect of an overexposed scene caused by
the low charge handling capacity of the HCG mode, which
leads to washed out details in the image.
of dB, as a function of exposure. When compared to a
traditional CMOS image sensor (with only low-CG), this
architecture shows an up to 5 dB increase in SNR under
low-light conditions, when high ISO-speed would be
utilized, by enabling the increased sensitivity and reduced
read noise of the HCG mode. In fact, the benefit of the
HCG mode can begin to be observed even at moderate
light levels . Higher SNR can be obtained in low-light
conditions, while the SNR, equal to that of a traditional
sensor, is retained at high-light conditions.
SNR vs. Exposure
Conditions: Max Analog Gain (8x), 5100K BB, 18% Reflectance Target
35
Figure 11: Zoomed-in portion of an overexposed image,
caused by HCG in a brightly lit scene.
Green Channel SNR (dB)
Figures 12 and 13 show the benefit of properly exposing this
scene with Aptina DR-Pix technology LCG mode enabled.
30
25
20
15
10
5
0.00001
0.001
0.001
0.01
Target Exposure (lux * sec)
DR-Pix
Figure 12: Properly exposed bright light scene, 2,430 lux
daylight illuminant, 125ms, enabled by the LCG mode of
Aptina’s MT9H004 sensor with Aptina DR-Pix technology.
Traditional CIS
Figure 14: The Aptina DR-Pix technology approach results
in much higher SNR at low-light conditions, compared to
a traditional pixel with only LCG.
To relate this to real world camera operation, Figures 15
and 16 show images taken from Aptina’s 16 megapixel
DSLR (MT9H004) sensor with Aptina DR-Pix technology at
ISO speeds of 100 and 12800.
Figure 13: Zoomed-in portion of a properly exposed
brightly lit scene, which retains detail in bright areas due to
large signal handling capacity.
To illustrate the benefit for camera applications, the sensor’s
measured parameters (quantum efficiency, full-well, read
noise, and conversion gain) are used to calculate SNR. The
illuminant spectrum used for the calculations is a 5100K
black body source.
The plot of Figure 14 shows the green channel SNR, in units
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved.9/10 EN.L
Figure 15: ISO 100 image captured by Aptina’s
16 megapixel DSLR sensor
An
ObjectiveDR-Pix
Look at
FSI and BSI
Leveraging
Technology
High conversion gain mode
Low conversion gain mode
References
1. M. Mase, S. Kawahito, M. Sasaki, S. Wakamori, “A Wide
Dynamic Range CMOS Image Sensor with Multiple
Exposure-Time Signal Outputs and 12-bit Column-Parallel
Cyclic A/D Converters,” IEEE J. of Solid State Circuits, Vol. 40,
NO. 12, pp 2787-2795, December 2005.
2. M. Sasaki,,M. Mase, S. Kawahito, S. Wakamori,” A Wide
Dynamic Range CMOS Image Sensor with Multiple ShortTime Exposures,” Proc. IEEE Sensors, Vlenna, Oct2004, pp.
967-972.
3. S. Kavadis, B. Dierickx, D. Scheffer, A. Alaets, D. Uwaerts,
and J. Bogaerts, “ A Logarithmic Response CMOS Image
Sensor with On-Chip Calibration,” IEEE J. Solid-State
Circuits, vol. 32, Feb 1997, pp. 1146-1152.
Figure 16: ISO 12800 image captured by Aptina’s 16
megapixel DSLR sensor showing 5 dB increase in SNR in
HCG mode
4. M. Loose, K. Meier, J. Schemmel, “A Self-Calibrating
Single Chip CMOS Camera with Logarithmic Response,”
IEEE J. Solid-State Circuits, vol. 36, April 2003, pp. 586-596.
Conclusion
For many applications, including surveillance cameras,
automotive sign recognition systems, or DSLR cameras, the
need to capture very bright scenes and very dark scenes
is a reality. To capture these scenes, image sensors are
typically optimized for one extreme at the price of degraded
performance for the other. In reality, these imaging
applications require large FW capacities as well as increased
sensitivity. However, higher sensitivity could limit full-well
capacity, the maximum achievable SNR, and the total sensor
DR.
Several approaches have been introduced for implementing
intra-scene HDR, but they do nothing to improve lowlight sensitivity or reduce noise. Another approach,
offered by Aptina targets inter-scene DR. Called Aptina
DR-Pix technology, the technique combines two modes
of operation in one pixel design – low CG for large charge
handling capacity in bright scenes and a high CG mode with
increased sensitivity and low read noise for low-light scenes,
providing tremendous benefit for mirrorless and high-end
DSLR, surveillance, and notebook/PC cameras, as well as
automotive imaging systems, where images/video must be
captured in extreme low-light conditions without sacrificing
performance in high-light conditions.
aptina.com
©2010 Aptina Imaging Corporation. All rights reserved. 9/10 EN.L
About Aptina
Aptina is a global provider of CMOS imaging solutions
with a growing portfolio of products that can be found in
all leading mobile phone and notebook computer brands
as well as a wide range of products for digital and video
cameras, surveillance, medical, automotive and industrial
applications, video conferencing, barcode scanners, toys,
and gaming. Aptina enables Imaging Everywhere™ and
continually drives innovation in the market as seen with
the introduction of the first 14MP CMOS image sensor
for point-and-shoot and hybrid cameras (MT9F001), and
the industry’s first 5MP SOC with ¼” format (MT9P111).
Privately held Aptina’s investors include Riverwood
Capital, TPG Capital and Micron Technology. For additional
information on Aptina and news on technology webcasts
visit www.aptina.com. Subscribe to the latest news from
Aptina by copying the Aptina RSS feed into your favorite
RSS reader.
©2010 Aptina Imaging Corporation. All rights reserved.
Information is subject to change without notice. Aptina,
the Aptina logo, A-Pix, DR-Pix, and Imaging Everywhere,
are trademarks of Aptina Imaging Corporation.